Makie.jl plots
This page shows an example of plotting MCMCChains.jl with Makie.jl. The example is meant to provide an useful basis to build upon. Let's define some random chain and load the required packages:
using MCMCChains
chns = Chains(randn(300, 5, 3), [:A, :B, :C, :D, :E])Chains MCMC chain (300×5×3 Array{Float64, 3}):
Iterations = 1:1:300
Number of chains = 3
Samples per chain = 300
parameters = A, B, C, D, E
Summary Statistics
parameters mean std naive_se mcse ess rhat
Symbol Float64 Float64 Float64 Float64 Float64 Float64
A 0.0324 1.0226 0.0341 0.0345 915.3818 1.0016
B 0.0080 0.9717 0.0324 0.0315 831.5448 1.0016
C 0.0038 1.0124 0.0337 0.0356 868.9161 0.9997
D -0.0481 0.9906 0.0330 0.0319 917.3192 1.0012
E 0.0060 0.9734 0.0324 0.0339 1039.1511 1.0003
Quantiles
parameters 2.5% 25.0% 50.0% 75.0% 97.5%
Symbol Float64 Float64 Float64 Float64 Float64
A -2.0496 -0.6946 0.0758 0.7028 2.1171
B -1.8296 -0.6719 0.0039 0.6825 1.9137
C -1.9518 -0.6502 -0.0133 0.6851 1.9428
D -2.0363 -0.6882 -0.0158 0.6090 1.8324
E -1.8552 -0.6468 0.0328 0.6693 1.8702
A basic way to visualize the chains is to show the drawn samples at each iteration. Colors depict different chains.
using CairoMakie
CairoMakie.activate!(; type="svg")
params = names(chns, :parameters)
n_chains = length(chains(chns))
n_samples = length(chns)
fig = Figure(; resolution=(1_000, 800))
for (i, param) in enumerate(params)
ax = Axis(fig[i, 1]; ylabel=string(param))
for chain in 1:n_chains
values = chns[:, param, chain]
lines!(ax, 1:n_samples, values; label=string(chain))
end
hideydecorations!(ax; label=false)
if i < length(params)
hidexdecorations!(ax; grid=false)
else
ax.xlabel = "Iteration"
end
end
figNext, we can add a second row of plots next to it which show the density estimate for these samples:
for (i, param) in enumerate(params)
ax = Axis(fig[i, 2]; ylabel=string(param))
for chain in 1:n_chains
values = chns[:, param, chain]
density!(ax, values; label=string(chain))
end
hideydecorations!(ax)
if i < length(params)
hidexdecorations!(ax; grid=false)
else
ax.xlabel = "Parameter estimate"
end
end
axes = [only(contents(fig[i, 2])) for i in 1:length(params)]
linkxaxes!(axes...)
figFinally, let's add a simple legend. Thanks to setting label above, this legend will have the right labels:
axislegend(first(axes))
fig